[Swlug] DIY Geiger counter

Sun Mar 9 12:11:51 UTC 2025

(I'm assuming that the only reason for sending your 1st email off-list is because the photos would have been too big file-size-wise for the list probably, so I'm replying here all together.)

At Z+0000=2025-03-08Sat21:51:23, Rhys Sage sent:
> I've just had a go with the sensor properly blacked out with the crystal inside the housing. I put 3V through he BC547-diode setup and got curious readings on my oscilloscope.
> Rhys Sage
> [...]
> [IMG_20250308_145726.jpg  image/jpeg (2088143 bytes)] 
> [IMG_20250308_145659.jpg  image/jpeg (1637045 bytes)] 

    Thanks for the photographs.  Okay, the breadboard in the 2nd one looks like it's going to be a problem for such tiny pulses.  The loop area between the transistor leads could be a lot smaller, let alone the loop area within the jumper wires.  If your crystal pulses really are tens of nanoseconds, then the same kinds of problems apply as for RF circuits.

    There's a really good channel on YouTube called W2AEW (aka. Alan Wolke), and one of his videos is about various prototyping methods that work better for radio frequencies.  It might be this one, but YouTube does not work well for me anymore since Google went to war with every YouTube downloader tool in existence (`youtube-dl`, Invidious, NewPipe, you-name-it):-
https://inv.nadeko.net/watch?v=kH110yjYZ2g 

    I've also been experimenting with wire-wrapping, and found a way to minimise loop area last year by sandwiching the bird's nest between a pair of copper-clad boards -- for a project that is on hold while I was learning all about configuring & running a Postfix email server, but I should get back to it soon.

    If you do manage to get it to work on a breadboard with such big loop areas, then I guess it would be that the short pulses are being distorted and spread-out, but you are still able to detect the spread-out pulses.  But then you have noise to think about, and again, you pick-up a lot more noise when your loop areas are bigger, especially the current loops on the unamplified side, because the noise gets amplified, so I do recommend trying to minimise your loop areas, even if by using twisted wire pairs where components need to be further apart for whatever reason.

    So like, I'm guessing your pair of crocodile clips go to the actual crystal somewhere off the photo.  The red one connects via a resistor to the middle lead of the 1st transistor, which I think is the base looking at the other transistor for context, its base fed by the orange jumper from the emitter of the 1st, and its emitter connected by a light-blue jumper to the far-side negative/ground rail, so these are NPN transistors.  At least I /think/ that is a negative rail, only your far-side positive rail is connected to your near-side negative rail.  That doesn't make it electrically wrong, it just means that I can't rely on the rail labels in the photo.  I'm guessing that the far-side rails are more likely to the ones that match their labels.

    I'm also not sure what is going on with the little yellow jumper connecting to the same node as the red croc clip which I thought was input.  Normally the collectors of both NPN transistors connect to positive, and you need a resistor protecting both of their bases.  Otherwise the amplification of the 1st BJT can damage the 2nd BJT, because base to emitter is effectively only a diode-drop to ground, so you need the resistor either between the transistors or on the collector of the 1st to prevent the short or overcurrent.

    Anyway, point is that if you connected the crystal via a twisted pair, that would reduce a huge amount of your loop area where it matters most, both in terms of reducing inductances so that you can detect pulses of very short duration, and to avoid receiving EMI noise due to the loop antenna effect.  If you are talking on the scale of tens of nanoseconds, then all of those loops on your breadboard are big enough to be pretty significant antennae, I think, so starting with the biggest ones and reducing them should see some improvements.

    As for the big floppy black jumper crossing from one supply rail to another, loop area in the supply can be truncated by adding capacitors across the DC supply as close to where the current changes as possible, because what matters is the area within the path of changing current.  The DC component of a current does not matter, and capacitors filter-out most of the ripple on a DC supply.  You could place a couple of capacitors on a pair of rails near where they are used, often a low ESR capacitor for the small ripple and maybe a bigger electrolytic for the big ripple, and this would reduce the width across the rails to tenth of an inch, providing that you only use the rails on one side of the breadboard.  This will help to reduce the loop area between the supply cap(s), the crystal, & the 1st transistor -- look at that 1st current loop and minimise its area.  Then look at the 2nd current loop area and try to get that down a bit too, and so on.

At Z+0000=2025-03-09Sun00:54:21, Rhys Sage via Swlug sent:
> I had a go at putting the Geiger counter together today. I'd been waiting while paint cured on the crystal housing. So I put two BC547s with a 10K resistor and 1uF capacitor, reading off the BPX61. I connected it to my oscilloscope and there is a detected variance but I can't see much happening on the graph.
> 
> I'm getting a Vmax of 10.74, Vmin of 6.82. The changes are probably too rapid for the scope to read as the frequency is 0khz.
> 
> I wonder if the BC547 is either not sensitive enough or whether the 1uf capacitor needs to be smaller.

    Okay, I've looked it up.  According to the 1987 Motorola Small-Signal Semiconductors databook, BC547 does indeed have a base-centric pinout and is NPN as I guessed from context -- I also see that it has a current-gain bandwidth product of 150MHz minimum or 300MHz typical.  That seems like a lot, but is it enough for a 20ns pulse?

    I would think of this by imagining a sine wave with a 40ns periodicity to begin with.  That sine wave would have a frequency of 25MHz.  Assuming the minimum current-gain bandwidth product of 150MHz, a 25MHz sine wave would be amplified with a current gain of 6 by the 1st transistor, and about 36 by both transistors.  Is that enough gain?  I don't know how much current the crystal outputs or how much you need on the output, but I guess it would be borderline good enough, especially when you consider that it will still be possible to detect a distorted and spread-out pulse.

    If you use an envelope detector on the input, using a fast Schottky diode (to "capture" the pulse), tiny-value capacitor (I'm thinking one of those little tan-coloured ceramic disk ones), and high-value resistor (to drain the small capacitor slowly), and spread each pulse to be, say, 200ns and that distortion to sine waves of wavelength no shorter than 400ns is acceptable, then you're looking at a sine wave frequency of 2.5MHZ, a current-gain in the 1st transistor of 60, and in both transistors a current-gain of 3600, which is much better -- but knock a zero off because you spread the energy of the pulse out by about tenfold, so you are starting from about a tenth of the current, I would have thought.  So 360.  Still a lot better than 36.

    I don't know enough about the crystal pulses to calculate suggested starting values for the "tiny-value capacitor" or "high-value resistor" in the envelope detector idea.

    It might be worth putting a 3rd transistor in the chain, but apparently if you are resorting to that, it is probably because you are getting something else wrong, and you might find that the circuit becomes too sensitive to noise.  I'm thinking that 2 of those transistors are probably just about enough, but the main problem is your loop areas.

> Anyway, a bit of progress. I might switch to the OPA2134PA chip and try that tomorrow.

    I'm thinking that this change will be insignificant compared to actually getting your signal path loop areas right down.  Current-gain bandwidth product more than about 300MHz is not going to help much here, I don't think.  Your transistors just about have that, but more likely all the loop areas are going to be adding a lot of inductance at this scale and picking-up a lot of small-signal noise as well.

    I think also check the wiring around the collector of that 1st transistor because it doesn't seem right to me, and protect the 2nd one with a resistor on its base to protect its base--emitter junction from overcurrent.  Or just move that resistor from the base of the 1st to the base of the 2nd, because actually it would protect both transistors if it was between them, assuming that the resistor is resistant enough (and assuming that your input voltage doesn't raise above a diode-drop above your positive supply voltage, which is not a problem here because of the tiny voltages and currents involved from the crystal).

    If you think you might have already shorted one or both, test them and replace them if their current-gain is off.  Their ordinary DC current gain is meant to be around 110 to 800 (it varies a lot with temperature, production batch, health of its silicon die, radiation (iirc.), & probably wind direction or which side of bed you got out of, but most multimeters should be able to give you a reading somewhere in that rough ballpark range).  The current-gain is really not a precise or stable value at the best of times -- I know this because I tried to design a current-limit circuit once that exploited this property and ended-up with a temperature run-off instability that defeated the object of the idea -- but at least if you see a reading that is really low or error you know the transistor is kaput.

    Correct me if I'm wrong, but as it stands in the photo, it looks like the 1st transistor has its base & collector connected together so is just acting as a diode, losing a diode drop and a little bit of attenuation, and then only the 2nd transistor is doing any amplification at all, which means that you definitely won't be getting enough amplification.  But before you fix this, put move the resistor to the base of the 2nd transistor so that you don't risk blowing it straightaway as soon as you get any significant signal on the input of the 1st.  They don't have integrated current-protection resistors so you have to be aware of this while you are using them and be sure to have a resistor somewhere in each current path.

    An alternative protection is to keep the resistor on the base of the 1st and add another one on the collector of the 1st as well (connecting to the positive supply rail), ensuring that the current coming out of its emitter cannot be too much for the base of the 2nd transistor.  This might be slightly better depending on circumstances because it minimises voltage drop if that matters here.

    If the amplification of both transistors is still not enough with having fixed these problems (huge loop areas causing inductance & noise, missing current protection resistor, 1st transistor collector connected to signal instead of supply), then adding 1 or 2 envelope detectors might be a useful way to spread out the pulses so that you can detect them, or even approximately count them using the analogue value of the accumulation and exponential decay of the envelope detector.  I'm not 100% sure whether it would be best to have an envelope detector before the amplification or after the amplification, or both -- both because you may need some spread beforehand, but not enough for the sample rate of your ADC / microcontroller, without losing too much voltage on the before side, and then be susceptible to more noise.  I think you could do with at least 1, but if there was a way to avoid having both, that would be ideal, but it would take some experimentation to know which side it is best to do it on.

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/20250309/bcdf5612/attachment.sig>