[Swlug] DIY Geiger counter

Tue Mar 11 08:22:22 UTC 2025

At Z+0000=2025-03-11Tue01:27:08, Alan Gray via Swlug sent:
> Hello Rhys,
> I think James's suggestions are on the right track, but I am not sure you're factoring in enough allowance for the frequencies involved.
> 
> The timings I gave suggest you are working in the range 0.3 to 50 MHz i.e.radio frequencies (RF). This is quite a specialised area within electronics.
> 
> Some of the capacitor values in the signal line feel quite high.

    I agree.  I suggested that he try one of the smallest caps that he has, which he said are 0.01µF or 10nF metal film ones.

> I would have expected pico value mica types.

    I was also thinking that 1nF or maybe smaller might be useful later when fine-tuning, but saw various other problems as bigger priorities, like getting the loop area right down.  But my suggestion of having a tight loop with a capacitor right at the photodiode would mean that the capacitor milks as much charge as possible from the photodiode during its tiny pulse of being "on", and then there is more time for this charge to flow into the rest of the circuit that has higher inductance due to the bigger loop areas and thus poor RF performance.  This might be enough of an improvement with the 10nF capacitance that he has to get something working that can later be refined.  The result would be a wider pulse spread-out in time, that a slower microcontroller has a better chance of not completely missing it between samples.

    I have this vague recollection that at the photon level, photodiodes (and LEDs) have a kind of one-to-one correspondence between photons and electrons, with a direct relation between the voltage drop and the photon energy|frequency|wavelength.  If only so many photons hit the photodiode, I would guess that only that same number of electrons can pass.  But if photons hit an unpowered photodiode, the freed electrons spontaneously recombine -- you lose the opportunity to let them pass.  So it still makes sense to "be ready" to let as many electrons pass as the photons allow, by having them stored on a nearby capacitor, effectively maximising the amount of charge that each pulse provides.  This charge will be more spread-out but there should be more of it overall, if you take the area under the curve, the integration of the current--time graph of the pulse, and by having more of both time and current of each pulse, I think it will be easier to detect it with a slow microcontroller with an analogue input.

    It should be possible to use this analogue value to calculate an estimated number of pulses of a particular type.  If the goal is to have enough resolution to count the individual pulses then definitely he should try 1nF or smaller values of capacitance across the photodiode, a faster microcontroller, and much tighter loops along the whole of the rest of the signal path.

> At those frequencies components have "unexpected" properties. For instance your common resistors also have inductance and capacitance. The values are low, but in this context they do matter. Old carbon rod types have low inductance, but modern film types commonly have a spiral cut in the film to adjust the resistance and give a tight tolerance. These have more inductance.
> 
> Much the same applies to capacitors. Electrolytic types often have a rolled internal structure and much inductance. Polystyrene block types are better than the polystyrene rolled ones, but probably not as good as ceramic disc types. Best are mica ones.

    The AM radio receivers and transmitter that I built as a child used the ceramic disc type, little tan-coloured flat balloons.  They are probably more than sufficient here.

    I've also used an axial type across IC chip power supply pins in high-speed digital circuits that I have been attempting to build more recently using high-speed CMOS (74HC) chips, aiming for tens or hundreds of megahertz.  These are also ceramic according to the datasheet I have for them: "Mono-Axial // Vishay // Dipped Axial Multilayer Ceramic Capacitors // for 50 - 100 Vdc".

    Maybe mica ones are only needed for gigahertz or tens of gigahertz ranges, for which I have no experience of DIY electronics.  I guess that they would not be needed here, but would do no harm either.

> The info is in manufacturers data, but unless you deal in RF you wouldn't normally look. After all it's just a simple passive component..isn't it.
> 
> Looking at your transistors. There are A, B and C variants around. C's are the highest gain. Are any of yours so marked? I think you are ok with the ones you are using, but keep in mind that RF types, like the components above, are constructed to have less capacitance.
> 
> I think you might find it useful to have a reliable test oscillator to work with. I did consider how you might build one, but then I thought you could just use a microcontroller clock. Essentially feed it into a potentiometer and tap off as much signal as you need. You'll probably need at least 1 MegaOhm connected between clock and ground so the loading has minimal effect. Start by using the scope to determine how much signal you are starting with and take it from there. Essentially you do not require a particular frequency or waveform. You simply need to be able to progressively turn down the level enough. The issue is likely to be getting it low enough.

    I'm wondering about making a test rig involving a 2nd photodiode and some other light-source, maybe an LED powered by a tiny capacitor or something -- the smaller the capacitor, the smaller the pulse.  Then the question is: How small of a pulse can you detect?  About 1 microsecond?  About 100 nanoseconds?  About 20 nanoseconds?

    Or drive the test LED & photodiode from your suggested test oscillator and determine where the frequency response loses power on the output of the amplifier.  Can it hear 1MHz?  Can it detect 5MHz?  What about 25MHz?

    If the scope is the limiting factor in terms of what frequency range it supports, it might be useful to make a little envelope detector on the output just to see if the test frequency is getting through.  The diode would not be needed because it's driven from an open-collector output, so just a capacitor there and a pull-up resistor, then see whether or by how much this test envelope voltage drops in response to different input frequencies.  That should allow to probe and feel about beyond the ranges of an oscilloscope, albeit with less directly-visual information.

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/20250311/71f0b7dc/attachment.sig>