[Swlug] DIY Geiger counter
James R. Haigh (+ML.LUG subaddress)
JRHaigh+ML.LUG at Runbox.com
Thu Mar 13 09:51:36 UTC 2025
[mostly written on 2025-03-12Wed afternoon/evening but fell asleep before sending; words like "yesterday", "today", etc. relative to Wednesday]
At Z+0000=2025-03-11Tue15:46:48, Rhys Sage via Swlug sent:
> I'm in the middle of my workday. Reading this all with great interest.
:-) I am also very interested in this project -- but that was interrupted yesterday [Tuesday] because my only mature tree, a roughly 15-year-old walnut tree mature enough to provide walnuts, suffered the blow of a chainsaw from a neighbour that thinks our boundary is 5 metres different to what I was told when I bought the paddock a couple of years ago. I was in a trance of emotional exhaustion most of last night, repeatedly falling asleep in the bath and waking up again with the bath water in my mouth. Crazy of me to be grieving a tree, but it's true. The walnut tree is partly why I bought the paddock. But owning the land of its roots clearly failed to protect it. I tried so hard to get it all right and it went wrong anyway. Such a pity.
Today I turn back to this thread because it has been a great source of excitement and worth to me. Sorry if I'm writing a bit too much most of the time, though.
> I played with chatGPT and of the list of transistors I have (I have some mosfets too),
Is this one of those exotic autocorrect mistakes? I think ChatGPT is an AI. What does it have to do with what transistors you have?
> it indicated my BC517s might be better than the BC547 setup. It'd certainly be simpler.
At least one advantage of the BC517 is that you have 1 less loop area to minimise, so that's good.
> The key points seem to be 30K times current gain and it's good up to 30v.
That's a huge current gain for a BJT. Okay, reading ahead, downloading the 344-page document, and seeing the schematic on page 78, I see already that the BC517 is labelled as "BC517 NPN Darlington Transistor" -- okay, so it is a Darlington Pair, and that explains its current gain being easily the square of a single BJT current gain.
The circuits that we were discussing involving the BC547s were just 1 connection different from being a Darlington Pair -- move the collector of the 1st transistor to join the open-collector output, the collector of the 2nd transistor. The current-limiting resistor that I suggested is then unnecessary, assuming that your output does not overdrive the 2nd transistor, but I was already making that assumption with the previous idea.
There is a slight catch with the Darlington Pair in that you do get a much bigger voltage drop on both the input and the output -- rather than being about 0.6v drop on the input and 0.2v drop on the output, you add another 0.6v of PN-drop to both, so about 1.2v and 0.8v, respectively, iirc.. This translates to increased inefficiency, and that is more of a problem if using it to drive a heavier load like LEDs or a motor if you're trying to do it efficiently (using switch-mode circuitry), but for small signals like this, it doesn't really matter, because you are only driving a high-impedance input anyway, so hardly any current flows across this roughly quadruple output voltage drop.
There's an alternative to the Darlington Pair called a Sziklai Pair, which solves the voltage drop problem, but I could not find any integrated ones when I searched for them a few years ago. But anyway, that doesn't matter here, as the extra voltage drop is not a problem given your chosen supply voltage.
Having accounted for that small topology change not having a significant effect on the function of this circuit, swapping the pair of discretes for an integrated Darlington Pair, is actually a very good idea, because it will make it easier to tighten the loop area of the current-change loops that flow through both of these transistors. Otherwise, the 2nd one would have to have a noticeably bigger loop area that the first, just because of the size of the discretes involved. Whereas, with both current loops deviating by only a fraction of the same silicon die, the 2nd one has negligible extra loop area than the 1st -- they are practically the same loop. And with a power supply capacitor adjacent to the integrated Darlington Pair, you can truncate both of these very similar current-change loops to be pretty small.
So not only does the BC517 integrated Darlington Pair make your circuit a little simpler to work with, it also provides another opportunity to minimise loop area of your loops of changing current -- any signal current is a changing current by nature of it being a signal conveying information, so any of the loops along your signal cascade whether it matters to have a high-bandwidth signal, need to be minimised, and this is one of them.
> I'll be using 4.7v. I'm only using 4.7v because that was the closest I could adjust my buck converter with the screwdriver at the weekend.
A couple of my previous calculations worked with 5v only because I was calculating a lower-bound value for resistance, so was giving some room for error by assuming an upper-bound voltage of 5v.
> I might have a 102 ceramic capacitor somewhere - there were some extra parts in a kit I built. Maybe I need to order a pack of polyester film capacitors? I have electrolytic and metal film assortment trays.
More on this in my reply to your 2nd email, because the schematic you have found has some key insights which affect this decision quite a lot.
> The photodiode in use is a BPX61. I did buy some BPX34s but couldn't find them. They're in a TO46 can style. Curiously no markings on them to indicate what they are nor the manufacturer's name. I'll attempt to include a photo.
Thanks for the photograph. I've seen these old-style cans before, but not with a round window like this. Nice to know about such things. They look heavy-duty, industrial quality. I like industrial-quality stuff -- as long as they are not inefficient or toxic, but I don't see any reason why a photodiode would be either of those.
I've seen LDRs in a roughly similar style, but flatter, don't remember any can enclosed around them, but they apparently have a heavy metal in them, mercury iirc., so not good. They have become obsolete in favour of photodiodes. I've not played with an LDR since secondary school. Photodiodes have a much better behaviour anyway.
At Z+0000=2025-03-11Tue23:59:35, Rhys Sage via Swlug sent:
> Reading around and researching this evening after my workday, I came across this document:https://nsi.ir/wp-content/uploads/2025/01/Radiation-Measurment.pdf
Got it. Heh, it's a publication of papers from a conference called "International Conference on Nuclear Science and Technology", which, dated 6th--8th May 2024, appears to have been hosted in Isfahan, Iran last year. Last year, a news story broke globally about a grave misuse of technology that Open-Source Hardware advocate Andrew Huang had given a talk about an aspect of its deployment (supply chain security) at a conference hosted in Israel a few years prior. I'm wondering whether countries host a conference to attract expertise that they are interested in using for very bad uses. I acknowledge that both sides of that war are doing atrocious things.
But it's really interesting that you've found a key part of the antidote (a circuit for a Geiger counter that can be used to help protect ourselves against nuclear fallout) -- but you've found this piece of antidote from a source (conference hosted by a state involved in war) that is almost definitely engaged in activities that are part of the problem for the risk of radioactive fallout. Curious.
Maybe I have misinterpreted this very complex geopolitical situation, but anyway, the important thing is that this document contains some very useful insight for understanding how to make a Geiger counter. Just stay away from page 345 as it is probably strictly forbidden in at least 47 countries.
> Page 78 has an interesting circuit diagram
This is really useful. At first it reminded me of XKCD 730 "Circuit Diagram" because of the strange, 3-pronged figure on the left side pointing "to centre of sun" (it even has the "blog cred" of an Arduino, lol!), but then I saw a mention of "envelopes" in the text, and I realised that this circuit actually has an envelope detector as I suggested, but after the amplifier, not before.
There's a really key insight here. The resistor of the envelope detector is 10MΩ which is even higher than I thought, especially on the amplified side. I had a sense that it would be high but like 10kΩ to 100kΩ. I have used resistors of 1MΩ or higher, but I'm less familiar with them, and had some problems with noise in the low-voltage circuits that used these, because the tiny currents that drew through them were less significant than the noise.
But they are suggesting a 10MΩ on the amplified side. That's really saying something. Firstly, whereas the other day I was umming and arring about whether to have an envelope detector before or after the amplifier or both, it is now clear to me that it's not going to work on the before side at all, because you would need resistance values on the order of 30k greater, hundreds of gigaohms, which I've not seen in low-voltage circuits, and I think you would have more of a noise problem just because you have more loops on the unamplified side. So that rules out before-side-only, and... [Actually no, it does not rule-out the idea of envelope detectors on both-sides -- it would still be possible to use the trick where you have a tight loop between photodiode and capacitor on the before-side to help improve the photodiode's efficiency of detecting very short pulses, but further reduce the decay rate of the envelope by also having the envelope detector on the amplified side. More on that below.]
Again, in a similar manner to my last schematic, the envelope detector's usual diode can be omitted because the current is only ever going to flow in one direction through the open-collector of the Darlington Pair. This capacitor charges when the Darlington Pair is active, and is slowly dissipated via the resistor in parallel with it as per the typical RC exponential decay.
This capacitor never reverse polarises, making it "electrolytic-compatible", so you could try one of your 0.1µF electrolytics here which is the same thing as the 100nF that this schematic suggests. However, it certainly does not have to be electrolytic; it can be any type that performs well at RF frequencies. I tend to prefer the nonelectrolytic types if they are available for a particular capacitance, for a few small reasons, like that electrolytics can fail by bulging or leaking, even outside of the capacitor plague, I think they still slightly more prone to failure than the others.
> that uses a Geiger-Muller tube.
Never heard of that. Presumably that is the strange 3-pronged symbol. I'm assuming that this can be replaced by your photodiode without issue.
> They're using a BC517 as the sole detector transistor. I think the 100n capacitor is slowing the pulses that will be received by an Arduino.
Sort of. The capacitor, the resistor, and the oneway behaviour of the Darlington Pair open-collector output as biased acts as an envelope detector, which, roughly speaking is drawing a dot-to-dot between peaks -- but doing so with line segments following an exponential decay curve. The overall effect results in a signal that has a much more modest bandwidth requirement, meaning that loop areas following that capacitor of the envelope detector need not be quite so tight -- depending on how frequently you want to sample at.
> The 10M resistor must be some kind of voltage pulse limiter.
It's part of the standard envelope detector circuit that I've been talking about. It forms an RC circuit with the capacitor, and allows it to drain slowly back to empty, with the usual exponential decay curve.
> I'm guessing wildly there - my actual electronics knowledge is about the level of a well read amoeba (or whatever Mr Pugh taught us in 5th form gen science back in 1982/1983). I'd love to know what those two components are really doing.
Did my explanations help? Envelope detector is also known as a peak detector. It's just about drawing an envelope around a series of waveform peaks. It's the simplest way to (approximately) demodulate an AM radio signal. In that case you have a carrier wave at ultrasonic frequency, and the envelope of it carries the sonic band of audible frequencies. Each peak raises the capacitor's charge, and between peaks, the resistor exponentially-decays it somewhat.
What we are trying to do here is to draw an envelope over the pulses. Each pulse charges the capacitor a little, and the resistor ensures that this charge exponentially-decays between pulses.
> So, that looks like something simple and worth trying on a breadboard. I have several spare breadboards so I can try that and compare with the circuits James sketched in unispace text.
Now I see the 10MΩ resistor value that they suggest, I think my sketch of the circuit that tried the envelope detector on the before side won't work, as you won't be able to find resistor values that work, and if you did, the circuit would probably be very susceptible to noise. I was not sure whether the limiting-factor for bandwidth would be the amplifier or the envelope detector. I initially thought that the passive components of the envelope detector would be best on the unamplified side where speed matters most, but this clue tells me that an envelope detector definitely needs to be on the amplified side.
So the envelope detector on the amplified side is definitely needed, but the idea of adding a tiny capacitor in a really tight loop with the photodiode may still be a valid boost to the circuit, especially when on a breadboard with a lot of loop areas. The page 78 circuit already has a resistor in series with the detector component, so my only suggested addition would be some tiny value of capacitance added across the photodiode. But then again, the parasitic capacitance of the photodiode might already be enough for the tiny amounts of charge per pulse, so even this might not be any use. It might still be worth trying, though. I suggest 1nF, but as you don't have that try the smallest you have, the 10nF metal film I think you said.
For the sake of experimentation and potential improvement, here's my revised version with the extra capacitor in, versus the page 78 version without it, and it will be interesting to know whether it helps at all (I've also just realised that the page 78 version lacks the biasing pull-up resistor R1 of previous schematic, which may improve analogue accuracy and response by steering clear of the lower voltage rail, so my variant retains this too):-
(view in any fixed-width font; the semicolons should align) ;
(if mangled by your email client then try "view source" as well) ;
+ + + ;
| | | ;
+ |+ + |+ > ;
| ---C | ---C < R ;
C .---o--. ---1 > ---3 > 3 ;
R |xxxxxx|+ | < R | < OPEN-COLLECTOR ;
Y:::PHOTO xx---C - > 1 | | AMPLIFIED ;
S:::DIODE xx---2 < (.--o----o-- OUTPUT (INVERTED) ;
T |xxxxxx| R2 | (C |) TO PULLED-UP ADC OR ;
A '------o------\/\/\--o----B C) CURRENT-LIMITED LED ;
L (E--B) ;
( E) ;
BC517 | ;
- ;
Minimise the area of the crystal--capacitor loop marked with cross-hatching.
Capacitors with polarity marked can be electrolytic, but don't have to be.
Component values:-
* C1 (all/any power supply capacitors) as high as practical, and as close as possible to where current changes, in order to truncate the loop area of the current-change loops.
* R1 probably quite high just for some DC bias, e.g. in the range 10kΩ to 10MΩ.
* R2 quite low mainly for some protective current-limitation, e.g. 1kΩ to 10kΩ.
* C2 very small -- try around 10nF to 1nF, maybe even smaller.
* C3 try 100nF (depends on your desired sample rate).
* R3 very high, e.g. 10MΩ (depends on your desired sample rate).
A lot of these are ranges because there are things that I don't know about the circuit (or some of them have a big range of tolerance anyway, like the R1 biasing resistor, because this just alters your zero point which you can account for later in software -- the reason for this is to avoid the nonlinearity and extra latency that you get near the rail voltages). Particularly I don't know the amount of charge in each pulse from the photodiode, and what your sample rate will be. The value for C3 and R3 given by the page 78 schematic is chosen for an Arduino Uno, so it makes sense to start with those values until you know otherwise. For faster or slower sample rates, they would need adjusting to give you more or less definition of your envelope.
Even with C2 in a tight loop with the sensing device, it's still quite important to minimise the loop area of the loop that goes from the nearest supply capacitor (C1), through C2 or the photodiode, through R2, through the BC517, and back to C1 (to reduce inductance and EMI pickup), and also to minimise the loop from C1, through C3, through the BC517, and back to C1 (to reduce inductance and EMI emission). If C2 is omitted, then it's especially important to minimise these next 2 loop areas. The loop areas after C3 should still be minimised, but with much less consequence, because the rate of change of current in these loops will be much slower anyway such that your microcontroller can sample the signal meaningfully.
So overall, my latest circuit is suggesting to keep the R1 biasing resistor from your existing BC547-based circuit, and add a couple of capacitors (C1 & C2) to boost performance by truncating the fastest-changing current paths into much tighter loops.
It's interesting that they are using a boost convertor to raise the voltage to 27vDC, because one of my earlier suggestions to get something working was to try a higher voltage, if your components support it. However, I don't think photodiodes need, or can even tolerate, that kind of voltage. Your circuit would also be more complex and less efficient for having the boost convertor included in it, so I hope it is found to be unnecessary. I reckon that by adding the 2 capacitors and the resistor that I suggest to improve the overall efficiency and performance of the detection, it's hopefully possible to steer clear of needing any voltage higher than 5v, which will also save the extra cost, bulk, and risk of the boost convertor.
Risk because boost convertors can fail in very destructive ways, producing voltages much higher than you designed for -- unlike buck convertors which are strictly limited by their supply voltage. Such convertors often have major design flaws. Did you know that on most cheap buck or boost convertors with a constant-current function, that constant-current function can fail very easily without any internal component failure, simply by shorting the input GND to the output GND -- there's actually a current shunt resistor on the ground side on most of them. In some projects, it is really difficult or inconvenient to prevent the input and output grounds from shorting. It's easy enough to design one with the shunt resistor on the high side to avoid this problem, but most bulk manufacturers or sellers don't care. I once modded an ordinary constant-voltage buck convertor to add a constant-current mode, and I'm pretty sure that I avoided this problem by adding the shunt to the high side.
Most are also sold with a claim of efficiency that is much higher than real because they didn't bother to account for the voltage drop of the diode, and just quote the efficiency of the main switching chip on the board. For low-voltage circuits like 3.3v or less, the diode drop has a big impact on efficiency. Synchronous switching promises to solve this problem, but with the market flooded with the false claims of the ones that rely on a flyback diode with its voltage drop, it can be difficult to find the actual good ones. I'm not very good at online shopping at the best of times; it's even more difficult when the listings are falsely advertised. If you find some good synchronous buck convertors that are definitely synchronous (with low-voltage efficiencies of around 96% or more, iirc.), I'd be interested to know as well.
Hopefully it's possible to avoid the boost convertor. Reducing loop area and adding the suggested capacitors and keeping the biasing resistor from previous attempt should all help improve sensitivity and performance, and hopefully be enough to work on 4.7v or so, in which case your project will be a little simpler, lighter, cheaper, & more efficient. For a battery-powered handheld, the better efficiency will translate directly into longer battery runtime.
It'll be interesting to see whether you notice much of a difference for adding/removing C2 or R1.
> This weekend is a 3-day weekend so I have a bit more time than usual even though Monday I have to take 'er indoors to one of her doctor's appointments.
It'll be exciting to see whether you get it working, or rather, /how/ you get it working. :-)
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: 2025-03-12Wed14:03:14+00_screenshot.png
Type: image/png
Size: 153863 bytes
Desc: not available
URL: <http://mailman.lug.org.uk/pipermail/swlug/attachments/20250313/340368fc/attachment-0001.png>
-------------- 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/20250313/340368fc/attachment-0001.sig>
More information about the Swlug
mailing list